❌

Normal view

Apollo Backs AI Startups in Bid to Bankroll Hardware Boom

16 September 2026 at 22:51

Apollo Global Management, known for leveraged buyouts and hefty financing deals, has started investing in AI and related hardware startups with the ambition to lend money to them in the future.

In one previously unreported deal, Apollo has already invested in the low tens of millions of dollars in Mercor’s latest funding round, which is set to value the data-labeling startup at $20 billion, according to people familiar with the matter. It has also invested in chip startup SiFive and in Hadrian, which says it’s building AI-powered factories, in transactions the companies disclosed earlier this year. It expects to make more investments in these categories, the people said.

Drone Delivery Startup Nears $20 Billion Valuation in New Funding Talks

16 September 2026 at 21:53

Zipline, a startup making drones that can deliver small packages like meals and medical supplies, is in talks to raise $1 billion at a valuation of roughly $20 billion, according to three people with knowledge of the fundraise.

The funding round, if it closes, would more than double its valuation from its last funding round, which valued the company at $7.6 billion after the investment. It follows a partnership announced last month with Uber to expand delivery by drones across the U.S. and expectations by investors that the costs of drone delivery are falling, expanding what was once a niche market.

How to Stay on Top of Tech Trends with Google Gemini

16 September 2026 at 16:36

Even for technology leaders, it can feel impossible to keep up with constant advances in the field.

This is especially true in the age of AI, with vendors racing to roll out model updates with new capabilities. Even if technology leaders had time to attend every conference and read every tech publication, it would be a full-time job to simply figure out which trends represent real transformation opportunities, and which are merely hype.

With Google Gemini, you can generate a tailored digest that summarizes the developments most likely to have an impact on your specific industry and organization. Armed with this up-to-the-minute information, you can keep your leadership and workforce apprised of market shifts and technology breakthroughs, protect your organization from expensive distractions, and make high-stakes procurement decisions with confidence.

Step 1: Generate Your Tech Forecast

Open up a new chat, and prompt Google Gemini to provide you with a summary of current technology trends. For a more tailored summary, provide as much information as possible about your company and industry. You might, for example, tell Gemini that you’re the CTO at a mid-sized regional commercial and retail bank in the Pacific Northwest, with a mix of legacy on-premises and modernized cloud infrastructure.

Here’s a sample prompt:

I am the CTO of [company] in [industry]. I want to understand emerging technology trends. Summarize the top five emerging technologies with the most significant potential impact on [industry]. For each technology, list its potential benefits and challenges, and suggest how it could impact [company] in the next two to three years.

Gemini will quickly generate a list of technologies, broken down by their potential benefits, challenges, and overall impact. For the regional bank, Gemini might suggest using agentic AI for continuous risk assessment, pointing out the opportunity to proactively surface lending opportunities and flag potential risks.

Step 2: Refine the Output

Gemini’s summary will be based on the information you provided. To generate a more custom tech forecast, add more information about your existing environment and current challenges.

For example: Gemini might suggest deploying API-led middleware to bridge the gap between your legacy core and your modern cloud app; but maybe you already have middleware running, and the problem is that its fragility is leading to downtime. By adding that context, you can help Gemini focus its output on your specific needs.

Go back and forth with Gemini, providing additional context and feedback until the tech trends summary is exactly what you’re looking for.

Step 3: Automate Delivery

If you have to manually enter a new prompt every time you want an update on tech trends, you might not be as consistent as you’d like. Instead, prompt Gemini once to automatically provide you with an update each day, week, or month.

These sorts of prompts work best when they are completely self-contained, rather than the result of a long, messy chat history. So, ask Gemini to look back at your conversation and synthesize all of your real-world feedback into a single master prompt. Then, paste that master prompt into a new chat and explicitly tell Gemini your desired delivery schedule.

Gemini’s Scheduled Actions feature will take over, allowing you to manage or pause the recurring prompt in your account settings.

Step 4: Move from Knowledge to Action

Staying on top of tech trends is only really useful insofar as it spurs action. As you peruse Gemini’s scheduled summaries, consider where digging deeper into specific topics may help you optimize your operations, mitigate emerging risks, or prepare your IT environment for the future. Try this prompt:

Recommend three areas where [my company] can take proactive steps to stay ahead of the curve on [specific areas].

The CTO of the regional bank might, for example, ask Gemini for specific tips on network-based fraud intelligence. In this step, the role of Gemini is shifting slightly, from industry analyst to strategic sounding board. Because Gemini already has detailed information about your organization’s challenges, goals, and existing environment, its insights will be grounded in your specific reality.

By moving from high-level summaries to tactical next steps, you can use your understanding of tech trends to guide your organization’s IT investments in a marketplace that never stops moving.

Is Anthropic Making Claude Too Human to Control?

16 September 2026 at 15:02

Anthropic’s call to slow down AI development to keep humans safe has dominated the news cycle and drew similar calls from the leaders of OpenAI, Microsoft and others (and subtle disagreement from Mark Zuckerberg). 

But Mustafa Suleyman, who runs Microsoft’s AI model development, has a warning about Anthropic itself. 

Suleyman, who co-founded DeepMind (now part of Google), says Anthropic’s approach of treating its Claude models as if they were sentient, similar to humans, could produce a dangerous outcome. He called out Anthropic’s Claude Constitution, a document that treats the model as an intelligent entity and offers guidelines for how it should handle ethical dilemmas. Suleyman said that by incorporating the document into its training, Anthropic could make Claude harder to control. 

Suleyman’s concerns emerged when Anthropic updated the constitution at the start of this year to leave room for the possibility that Claude has some sort of “consciousness” or “moral status,” a change from the original document.

Copper ion-induced MYH9 polymerization executes cuproptosis

16 September 2026 at 14:28

Publication date: Available online 14 September 2026

Source: Cell

Author(s): Liu-zheng Wu, Xiang-yu Mi, Zai-jun Liu, Yang Ding, Wen-bin Hong, Chen-xi Cai, Hai-jing Hu, Yuan-li Ai, Bi-Xing Zhao, Duo-duo Qiu, Jing-song Ma, Xue-qin Chen, Han You, Tianwei Lin, Fu-nan Li, Xue-hui Hong, Hang-zi Chen, Qiao Wu

Glycocalyx-edited mesenchymal stem/stromal cell therapy in advanced osteoporosis

16 September 2026 at 14:28

Publication date: Available online 11 September 2026

Source: Cell

Author(s): Jose M. Moraleda, Luis F. Linares, David García-Bernal, Nuria Lozano-Rivas, Jenifer Ruiz-Sará, Miguel Blanquer, M. Dolores López-Lucas, Darío Sánchez-Salinas, M. Carmen Algueró, M. Mar Molina-Molina, Francisco Ruiz-Espejo, José L. Peris-Serra, Pilar M. Arrabal, José Becerra, Jonathan Kiwitt-Cárdenas, Carlos Marras, Ana M. García-Hernández, Robert Sackstein

Mitochondrial ATP promotes T cell differentiation through chromatin accessibility

16 September 2026 at 14:28

Publication date: Available online 11 September 2026

Source: Cell

Author(s): Charles Ng, Tak Shun Fung, Dayi Li, Korbinian N. Kropp, Luis F. Somarribas Patterson, Alexandria Markovitz, Daniel N. Weinberg, Olivia Jones, Ji-Young Kim, Guoan Zhang, Richard Koche, Mara Monetti, Huayuan Tang, Yun He, Zhengshuang Xu, Xin Cai, Ziqi Yu, Geetha Bhagavatula, Sean P. Colgan, Ya-Hui Lin

Pourquoi il est si difficile de déterminer l'âge d'un champignon ?

16 September 2026 at 14:22
Comment définir les contours d'un champignon et estimer sa longévité ? Des questions qui semblent simples, mais auxquelles la science peine encore à répondre. Rencontre avec Ayla Mongès, chercheuse en écologie fonctionnelle, qui tente de percer les mystères de ces organismes hors normes.

Nuclear fusion testbed gets 50-MW pulsed-power upgrade to push core plasma temperatures

16 September 2026 at 14:22

Engineers at compact fusion company nT-Tao have brought a 50-megawatt pulsed-power subsystem into active service. The hardware delivers high-energy bursts to the company’s C3 prototype, a machine designed to test magnetic confinement in dense plasma regimes.

This installation marks a critical step toward exceeding 1.5 million degrees Celsius, the highest thermal benchmark recorded on the platform to date.

Injecting large bursts of electricity into an ionized gas triggers severe circuit instability. As the plasma heats up during each pulse, its internal electrical properties fluctuate within fractions of a millisecond. These rapid swings alter the electrical impedance of both the plasma and the delivery hardware.

“The power must also be controlled precisely as the electrical characteristics of the system and the plasma change rapidly during each pulse,” said the company in a press release.

If the supply fails to match these shifts instantly, electrical resonance breaks down and the gas reflects energy back into the driving circuits. To prevent this loss, nT-Tao paired the 50-megawatt unit with a proprietary real-time nonlinear control system.

Nonlinear real-time tuning

The controller tracks circuit changes during the pulse and updates operating parameters continuously. This active tuning maintains resonance throughout the discharge, ensuring maximal energy transfers into the plasma core under fast-varying conditions.

CTO and co-founder Boaz Weinfeld pointed out that generating 50 megawatts is relatively straightforward compared to the task of coordinating 50 separate modules in real time. The primary engineering accomplishment lies in forcing multiple units to function as one synchronized circuit.

The physical power supply expands the company’s Modular Energy Generator Architecture, known as MEGA. The setup distributes the power load across 50 compact units called Power Electronic Generators (PEGs), rather than relying on a single, large power source. Each PEG produces approximately 1 megawatt of output.

Technicians grouped these generators into ten parallel branches, placing five parallel-wired modules on each branch. The complete network operates across switching frequencies between 50 and 2,000 kilohertz.

“It represents a roughly five-fold scale-up from an earlier generation that integrated 12 PEGs into an approximately 10 MW system,” added the company.

Hardware scalability and recent milestones

Using 1-megawatt modules provides practical manufacturing and operational advantages. Technicians can connect modules in series or parallel arrangements to meet specific circuit requirements. According to Weinfeld, fusion systems that cannot be produced at scale will fail commercially. Standardized, repeatable modules keep the physical footprint compact and allow capacity growth without redesigning core hardware.

The 50-megawatt power feed supports a broader strategy to turn plasma physics theory into viable, compact reactor hardware. Assembly took two months, with the machine reaching operational status and firing initial plasma pulses in January.

That testbed followed an earlier experimental series with the C2-A machine, which registered plasma temperatures around 100 electronvolts, or about 10^6 degrees Celsius.

By combining modular power electronics with high-speed nonlinear controls, the team has established the electrical drive required to push plasma past those early milestones toward commercial reactor temperatures.

“Together, the 50 PEGs and the real-time controller form the power architecture subsystem intended to deliver large amounts of energy to the plasma in precisely controlled pulses, moving nT-Tao closer to its next temperature milestone on the path toward a commercially viable compact fusion system,” concluded the company.

❌